China’s generative AI racing is heating up – not just with model development, but also with aggressive price wars. These are aimed at attracting enterprise users and developers. On April 8, the emerging AI startup DeepSeek slashed API usage rates for its large language model “DeepSeek–R1” by as much as 75%– undercutting competitors such Tencent, Baidu and iFlytek. According to public documentation,
DeepSeek’s new pricing places its model inference cost significantly below market averages. Tokens are priced at less $0.00007 Y=0.0005 (Wen Xin Yi Yan) per call. Baidu made its flagshipERNIE Bot (Wen Xin Yi Yan) completely free on April 1, including previously paid premium tiers, in an effort to boost its developer community and user base.
The major players are now competing on not only model capabilities, but also cost, developer experience, ecosystem integration, and more. Tencent Cloud has recently lowered the pricing of its “Hunyuan”model APIs. iFlytek, on the other hand, has bundled AI model services into education and healthcare platforms in order to increase adoption.
The price-cutting wave is similar to the cloud computing trends of the past decade. Hyperscalers such as AWS and Alibaba Cloud have lowered prices in order to scale usage and lock-in long-term customers. In China’s AI market, this reflects a shift in strategy from hype to monetization, as firms try to convert massive R&D investment into enterprise customers.
Analysts warn that the pricing war may squeeze margins of smaller players, while favored those with cloud infrastructure or corporate backing. DeepSeek, a spin-off from a top academic laboratory, has received funding from several Chinese tech venture capitalists. The company is betting on scale to bring sustainability.
As enterprises experiment with LLM based services, affordability could become the next major difference — especially in markets such as China, where cost efficiency often drives adoption over brand prestige.
